diff --git a/src/TensorFlowNET.Core/APIs/tf.math.cs b/src/TensorFlowNET.Core/APIs/tf.math.cs
index 042ab952..ce6dc4d6 100644
--- a/src/TensorFlowNET.Core/APIs/tf.math.cs
+++ b/src/TensorFlowNET.Core/APIs/tf.math.cs
@@ -21,6 +21,9 @@ namespace Tensorflow
         public MathApi math { get; } = new MathApi();
         public class MathApi
         {
+            public Tensor argmax(Tensor input, Axis axis = null, string name = null, int? dimension = null, TF_DataType output_type = TF_DataType.TF_INT64)
+                => gen_math_ops.arg_max(input, axis, name: name, output_type: output_type);
+
             public Tensor log(Tensor x, string name = null)
                 => gen_math_ops.log(x, name);
 
@@ -539,15 +542,12 @@ namespace Tensorflow
         public Tensor round(Tensor x, string name = null)
             => gen_math_ops.round(x, name: name);
 
-        public Tensor cast(Tensor x, TF_DataType dtype = TF_DataType.DtInvalid, string name = null)
+        public Tensor cast(Tensor x, TF_DataType dtype, string name = null)
             => math_ops.cast(x, dtype, name);
 
         public Tensor cumsum(Tensor x, int axis = 0, bool exclusive = false, bool reverse = false, string name = null)
             => math_ops.cumsum(x, axis: axis, exclusive: exclusive, reverse: reverse, name: name);
 
-        public Tensor argmax(Tensor input, int axis = -1, string name = null, int? dimension = null, TF_DataType output_type = TF_DataType.TF_INT64)
-            => gen_math_ops.arg_max(input, axis, name: name, output_type: output_type);
-
         public Tensor square(Tensor x, string name = null)
             => gen_math_ops.square(x, name: name);
         public Tensor squared_difference(Tensor x, Tensor y, string name = null)
diff --git a/src/TensorFlowNET.Core/Binding.Util.cs b/src/TensorFlowNET.Core/Binding.Util.cs
index c79b6f3a..8a744543 100644
--- a/src/TensorFlowNET.Core/Binding.Util.cs
+++ b/src/TensorFlowNET.Core/Binding.Util.cs
@@ -549,6 +549,8 @@ namespace Tensorflow
                     return tensors.dtype;
                 case IEnumerable<Tensor> tensors:
                     return tensors.First().dtype;
+                case RefVariable variable:
+                    return variable.dtype;
                 case ResourceVariable variable:
                     return variable.dtype;
                 default:
diff --git a/src/TensorFlowNET.Core/Contexts/ExecuteOpArgs.cs b/src/TensorFlowNET.Core/Contexts/ExecuteOpArgs.cs
index ecdcff8e..8710ea5d 100644
--- a/src/TensorFlowNET.Core/Contexts/ExecuteOpArgs.cs
+++ b/src/TensorFlowNET.Core/Contexts/ExecuteOpArgs.cs
@@ -1,5 +1,6 @@
 using System;
 using System.Collections.Generic;
+using System.Diagnostics;
 using System.Text;
 using static Tensorflow.Binding;
 
@@ -11,11 +12,13 @@ namespace Tensorflow
         public object[] OpInputArgs { get; set; }
         public Dictionary<string, object> OpAttrs { get; set; }
         
+        [DebuggerStepThrough]
         public ExecuteOpArgs(params object[] inputArgs)
         {
             OpInputArgs = inputArgs;
         }
 
+        [DebuggerStepThrough]
         public ExecuteOpArgs SetAttributes(object attrs)
         {
             OpAttrs = ConvertToDict(attrs);
diff --git a/src/TensorFlowNET.Core/NumPy/Axis.cs b/src/TensorFlowNET.Core/NumPy/Axis.cs
index 3c43686d..05644640 100644
--- a/src/TensorFlowNET.Core/NumPy/Axis.cs
+++ b/src/TensorFlowNET.Core/NumPy/Axis.cs
@@ -56,7 +56,7 @@ namespace Tensorflow
             => constant_op.constant(axis);
 
         public override string ToString()
-            => $"({string.Join(", ", axis)})";
+            => IsScalar ? $"{axis[0]}" : $"({string.Join(", ", axis)})";
     }
 }
 
diff --git a/src/TensorFlowNET.Core/NumPy/Implementation/RandomizedImpl.cs b/src/TensorFlowNET.Core/NumPy/Implementation/RandomizedImpl.cs
new file mode 100644
index 00000000..244fc61b
--- /dev/null
+++ b/src/TensorFlowNET.Core/NumPy/Implementation/RandomizedImpl.cs
@@ -0,0 +1,29 @@
+using System;
+using System.Collections.Generic;
+using System.Runtime.InteropServices;
+using System.Text;
+
+namespace Tensorflow.NumPy
+{
+    public class RandomizedImpl
+    {
+        [AutoNumPy]
+        public NDArray permutation(int x) => new NDArray(random_ops.random_shuffle(math_ops.range(0, x)));
+
+        [AutoNumPy]
+        public NDArray permutation(NDArray x) => new NDArray(random_ops.random_shuffle(x));
+
+        [AutoNumPy]
+        public void shuffle(NDArray x)
+        {
+            var y = random_ops.random_shuffle(x);
+            Marshal.Copy(y.BufferToArray(), 0, x.TensorDataPointer, (int)x.bytesize);
+        }
+
+        public NDArray rand(params int[] shape)
+            => throw new NotImplementedException("");
+
+        public NDArray randint(long x)
+            => throw new NotImplementedException("");
+    }
+}
diff --git a/src/TensorFlowNET.Core/NumPy/NDArray.Index.cs b/src/TensorFlowNET.Core/NumPy/NDArray.Index.cs
index 5e36c48c..160e1d6e 100644
--- a/src/TensorFlowNET.Core/NumPy/NDArray.Index.cs
+++ b/src/TensorFlowNET.Core/NumPy/NDArray.Index.cs
@@ -45,6 +45,8 @@ namespace Tensorflow.NumPy
             {
                 if(mask.dtype == TF_DataType.TF_INT32)
                     return GetData(mask.ToArray<int>());
+                else if (mask.dtype == TF_DataType.TF_INT64)
+                    return GetData(mask.ToArray<long>().Select(x => Convert.ToInt32(x)).ToArray());
 
                 throw new NotImplementedException("");
             }
diff --git a/src/TensorFlowNET.Core/NumPy/NumPy.Sorting.Searching.Counting.cs b/src/TensorFlowNET.Core/NumPy/NumPy.Sorting.Searching.Counting.cs
index c657294e..b9ad9812 100644
--- a/src/TensorFlowNET.Core/NumPy/NumPy.Sorting.Searching.Counting.cs
+++ b/src/TensorFlowNET.Core/NumPy/NumPy.Sorting.Searching.Counting.cs
@@ -17,7 +17,13 @@ namespace Tensorflow.NumPy
             => new NDArray(math_ops.argmax(a, axis ?? -1));
 
         [AutoNumPy]
-        public static NDArray unique(NDArray a)
-            => throw new NotImplementedException("");
+        public static (NDArray, NDArray) unique(NDArray a)
+        {
+            var(u, indice) = array_ops.unique(a);
+            return (new NDArray(u), new NDArray(indice));
+        }
+
+        [AutoNumPy]
+        public static void shuffle(NDArray x) => np.random.shuffle(x);
     }
 }
diff --git a/src/TensorFlowNET.Core/NumPy/NumPy.Statistics.cs b/src/TensorFlowNET.Core/NumPy/NumPy.Statistics.cs
index aa7f8d67..806d38b2 100644
--- a/src/TensorFlowNET.Core/NumPy/NumPy.Statistics.cs
+++ b/src/TensorFlowNET.Core/NumPy/NumPy.Statistics.cs
@@ -13,6 +13,6 @@ namespace Tensorflow.NumPy
         public static NDArray amin(NDArray x, int axis = 0) => new NDArray(tf.arg_min(x, axis));
 
         [AutoNumPy]
-        public static NDArray amax(NDArray x, int axis = 0) => new NDArray(tf.arg_max(x, axis));
+        public static NDArray amax(NDArray x, int axis = 0) => new NDArray(tf.math.argmax(x, axis));
     }
 }
diff --git a/src/TensorFlowNET.Core/Numpy/NDArray.Creation.cs b/src/TensorFlowNET.Core/Numpy/NDArray.Creation.cs
index 87658a32..9b539a07 100644
--- a/src/TensorFlowNET.Core/Numpy/NDArray.Creation.cs
+++ b/src/TensorFlowNET.Core/Numpy/NDArray.Creation.cs
@@ -25,6 +25,9 @@ namespace Tensorflow.NumPy
         public NDArray(byte[] bytes, Shape shape, TF_DataType dtype) 
             : base(bytes, shape, dtype) { NewEagerTensorHandle(); }
 
+        public NDArray(long[] value, Shape? shape = null)
+            : base(value, shape) { NewEagerTensorHandle(); }
+
         public NDArray(IntPtr address, Shape shape, TF_DataType dtype) 
             : base(address, shape, dtype) { NewEagerTensorHandle(); }
 
diff --git a/src/TensorFlowNET.Core/Numpy/NDArray.cs b/src/TensorFlowNET.Core/Numpy/NDArray.cs
index 3f90db00..1adf7c0f 100644
--- a/src/TensorFlowNET.Core/Numpy/NDArray.cs
+++ b/src/TensorFlowNET.Core/Numpy/NDArray.cs
@@ -42,11 +42,9 @@ namespace Tensorflow.NumPy
         public NDArray reshape(Shape newshape) => new NDArray(tf.reshape(this, newshape));
         public NDArray astype(TF_DataType dtype) => new NDArray(math_ops.cast(this, dtype));
         public NDArray ravel() => throw new NotImplementedException("");
-        public void shuffle(NDArray nd) => throw new NotImplementedException("");
+        public void shuffle(NDArray nd) => np.random.shuffle(nd);
         public Array ToMuliDimArray<T>() => throw new NotImplementedException("");
         public byte[] ToByteArray() => BufferToArray();
-        public static string[] AsStringArray(NDArray arr) => throw new NotImplementedException("");
-
         public override string ToString() => NDArrayRender.ToString(this);
     }
 }
diff --git a/src/TensorFlowNET.Core/Numpy/Numpy.cs b/src/TensorFlowNET.Core/Numpy/Numpy.cs
index 76a00418..89077796 100644
--- a/src/TensorFlowNET.Core/Numpy/Numpy.cs
+++ b/src/TensorFlowNET.Core/Numpy/Numpy.cs
@@ -75,28 +75,7 @@ namespace Tensorflow.NumPy
         public static bool allclose(NDArray a, NDArray b, double rtol = 1.0E-5, double atol = 1.0E-8,
             bool equal_nan = false) => throw new NotImplementedException("");
 
-        public static class random
-        {
-            public static NDArray permutation(int x)
-            {
-                throw new NotImplementedException("");
-            }
-
-            public static void shuffle(NDArray nd)
-            {
-
-            }
-
-            public static NDArray rand(params int[] shape)
-                => throw new NotImplementedException("");
-
-            public static NDArray randint(long x)
-                => throw new NotImplementedException("");
-
-            public static NDArray RandomState(int x)
-                => throw new NotImplementedException("");
-        }
-
+        public static RandomizedImpl random = new RandomizedImpl();
         public static LinearAlgebraImpl linalg = new LinearAlgebraImpl();
     }
 }
diff --git a/src/TensorFlowNET.Core/Sessions/BaseSession.cs b/src/TensorFlowNET.Core/Sessions/BaseSession.cs
index 46d62d82..3c994a6e 100644
--- a/src/TensorFlowNET.Core/Sessions/BaseSession.cs
+++ b/src/TensorFlowNET.Core/Sessions/BaseSession.cs
@@ -193,6 +193,9 @@ namespace Tensorflow
                         case double v:
                             feeds[i++] = new KeyValuePair<TF_Output, Tensor>(key._as_tf_output(), new Tensor(v));
                             break;
+                        case string v:
+                            feeds[i++] = new KeyValuePair<TF_Output, Tensor>(key._as_tf_output(), new Tensor(v));
+                            break;
                         case Array v:
                             feeds[i++] = new KeyValuePair<TF_Output, Tensor>(key._as_tf_output(), new Tensor(v, v.GetShape()));
                             break;
diff --git a/src/TensorFlowNET.Core/Tensors/constant_op.cs b/src/TensorFlowNET.Core/Tensors/constant_op.cs
index b55563f2..0dccb955 100644
--- a/src/TensorFlowNET.Core/Tensors/constant_op.cs
+++ b/src/TensorFlowNET.Core/Tensors/constant_op.cs
@@ -117,7 +117,7 @@ namespace Tensorflow
                 case Shape val:
                     return new EagerTensor(val.dims, new Shape(val.ndim));
                 case Axis val:
-                    return new EagerTensor(val.axis, new Shape(val.size));
+                    return new EagerTensor(val.axis, val.IsScalar ? Shape.Scalar : new Shape(val.size));
                 case string val:
                     return new EagerTensor(new[] { val }, Shape.Scalar);
                 case string[] val:
diff --git a/src/TensorFlowNET.Core/Tensors/tensor_util.cs b/src/TensorFlowNET.Core/Tensors/tensor_util.cs
index 4fa4d773..243b73d3 100644
--- a/src/TensorFlowNET.Core/Tensors/tensor_util.cs
+++ b/src/TensorFlowNET.Core/Tensors/tensor_util.cs
@@ -141,7 +141,24 @@ namespace Tensorflow
                 byte[] bytes = nd.ToByteArray();
                 tensor_proto.TensorContent = Google.Protobuf.ByteString.CopyFrom(bytes);
             }
-            else if (!values.GetType().IsArray)
+            else if (dtype == TF_DataType.TF_STRING && !(values is NDArray))
+            {
+                if (values is string str)
+                    tensor_proto.StringVal.Add(Google.Protobuf.ByteString.CopyFromUtf8(str));
+                else if (values is string[] str_values)
+                    tensor_proto.StringVal.AddRange(str_values.Select(x => Google.Protobuf.ByteString.CopyFromUtf8(x)));
+                else if (values is byte[] byte_values)
+                    tensor_proto.TensorContent = Google.Protobuf.ByteString.CopyFrom(byte_values);
+            }
+            else if (values is Array array)
+            {
+                // array
+                var len = dtype.get_datatype_size() * (int)shape.size;
+                byte[] bytes = new byte[len];
+                System.Buffer.BlockCopy(array, 0, bytes, 0, len);
+                tensor_proto.TensorContent = Google.Protobuf.ByteString.CopyFrom(bytes);
+            }
+            else
             {
                 switch (values)
                 {
@@ -166,32 +183,10 @@ namespace Tensorflow
                     case double val:
                         tensor_proto.DoubleVal.AddRange(new[] { val });
                         break;
-                    case string val:
-                        tensor_proto.StringVal.AddRange(val.Select(x => Google.Protobuf.ByteString.CopyFromUtf8(x.ToString())));
-                        break;
                     default:
                         throw new Exception("make_tensor_proto Not Implemented");
                 }
             }
-            else if (dtype == TF_DataType.TF_STRING && !(values is NDArray))
-            {
-                if (values is string str)
-                {
-                    tensor_proto.StringVal.Add(Google.Protobuf.ByteString.CopyFromUtf8(str));
-                }
-                else if (values is string[] str_values)
-                    tensor_proto.StringVal.AddRange(str_values.Select(x => Google.Protobuf.ByteString.CopyFromUtf8(x)));
-                else if (values is byte[] byte_values)
-                    tensor_proto.TensorContent = Google.Protobuf.ByteString.CopyFrom(byte_values);
-            }
-            else if (values is Array array)
-            {
-                // array
-                var len = dtype.get_datatype_size() * (int)shape.size;
-                byte[] bytes = new byte[len];
-                System.Buffer.BlockCopy(array, 0, bytes, 0, len);
-                tensor_proto.TensorContent = Google.Protobuf.ByteString.CopyFrom(bytes);
-            }
 
             return tensor_proto;
         }
diff --git a/src/TensorFlowNET.Core/Training/Saving/Saver.cs b/src/TensorFlowNET.Core/Training/Saving/Saver.cs
index c326267f..6138dba4 100644
--- a/src/TensorFlowNET.Core/Training/Saving/Saver.cs
+++ b/src/TensorFlowNET.Core/Training/Saving/Saver.cs
@@ -193,7 +193,7 @@ namespace Tensorflow
 
                 if (write_state)
                 {
-                    var path = NDArray.AsStringArray(model_checkpoint_path[0])[0];
+                    var path = model_checkpoint_path[0].StringData()[0];
                     _RecordLastCheckpoint(path);
                     checkpoint_management.update_checkpoint_state_internal(
                         save_dir: save_path_parent,
@@ -211,7 +211,7 @@ namespace Tensorflow
                 export_meta_graph(meta_graph_filename, strip_default_attrs: strip_default_attrs, save_debug_info: save_debug_info);
             }
 
-            return _is_empty ? string.Empty : NDArray.AsStringArray(model_checkpoint_path[0])[0];
+            return _is_empty ? string.Empty : model_checkpoint_path[0].StringData()[0];
         }
 
         public (Saver, object) import_meta_graph(string meta_graph_or_file,
diff --git a/src/TensorFlowNET.Core/ops.cs b/src/TensorFlowNET.Core/ops.cs
index 2e0c04e0..f499574f 100644
--- a/src/TensorFlowNET.Core/ops.cs
+++ b/src/TensorFlowNET.Core/ops.cs
@@ -165,8 +165,8 @@ namespace Tensorflow
             if (dtype == TF_DataType.TF_STRING)
                 return ret;
 
-            if (dtype != ret.dtype)
-                ret = gen_math_ops.cast(ret, dtype.as_base_dtype(), name: name);
+            if (dtype.as_base_dtype() != ret.dtype.as_base_dtype())
+                ret = gen_math_ops.cast(ret, dtype, name: name);
 
             return ret;
         }
diff --git a/src/TensorFlowNET.Keras/Preprocessings/DatasetUtils.index_directory.cs b/src/TensorFlowNET.Keras/Preprocessings/DatasetUtils.index_directory.cs
index e406176d..cfcbc4bc 100644
--- a/src/TensorFlowNET.Keras/Preprocessings/DatasetUtils.index_directory.cs
+++ b/src/TensorFlowNET.Keras/Preprocessings/DatasetUtils.index_directory.cs
@@ -3,6 +3,7 @@ using System;
 using System.Collections.Generic;
 using System.IO;
 using System.Linq;
+using static Tensorflow.Binding;
 
 namespace Tensorflow.Keras.Preprocessings
 {
@@ -50,8 +51,8 @@ namespace Tensorflow.Keras.Preprocessings
                 if (!seed.HasValue)
                     seed = np.random.randint((long)1e6);
                 var random_index = np.arange(label_list.Count);
-                var rng = np.random.RandomState(seed.Value);
-                rng.shuffle(random_index);
+                tf.set_random_seed(seed.Value);
+                np.random.shuffle(random_index);
                 var index = random_index.ToArray<int>();
 
                 for (int i = 0; i < label_list.Count; i++)
diff --git a/src/TensorFlowNET.Keras/Preprocessings/Preprocessing.image_dataset_from_directory.cs b/src/TensorFlowNET.Keras/Preprocessings/Preprocessing.image_dataset_from_directory.cs
index f5b52dfb..fa19987b 100644
--- a/src/TensorFlowNET.Keras/Preprocessings/Preprocessing.image_dataset_from_directory.cs
+++ b/src/TensorFlowNET.Keras/Preprocessings/Preprocessing.image_dataset_from_directory.cs
@@ -115,8 +115,8 @@ namespace Tensorflow.Keras
             var start_positions = np.arange(0, num_seqs, sequence_stride);
             if (shuffle)
             {
-                var rng = np.random.RandomState(seed);
-                rng.shuffle(start_positions);
+                tf.set_random_seed(seed);
+                np.random.shuffle(start_positions);
             }
 
             var sequence_length_tensor = constant_op.constant(sequence_length, dtype: index_dtype);
diff --git a/test/TensorFlowNET.UnitTest/NumPy/Randomize.Test.cs b/test/TensorFlowNET.UnitTest/NumPy/Randomize.Test.cs
new file mode 100644
index 00000000..38a4fbbe
--- /dev/null
+++ b/test/TensorFlowNET.UnitTest/NumPy/Randomize.Test.cs
@@ -0,0 +1,27 @@
+using Microsoft.VisualStudio.TestTools.UnitTesting;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using Tensorflow;
+using Tensorflow.NumPy;
+
+namespace TensorFlowNET.UnitTest.NumPy
+{
+    /// <summary>
+    /// https://numpy.org/doc/1.20/reference/random/index.html
+    /// </summary>
+    [TestClass]
+    public class RandomizeTest : EagerModeTestBase
+    {
+        [TestMethod]
+        public void permutation()
+        {
+            var x = np.random.permutation(10);
+            Assert.AreEqual(x.shape, 10);
+            var y = np.random.permutation(x);
+            Assert.AreEqual(x.shape, 10);
+            Assert.AreNotEqual(x.ToArray<int>(), y.ToArray<int>());
+        }
+    }
+}